Zylor Systems
All work

AI & EdTech

SignLearn — AI-Powered Sign Language Learning for Deaf Children

An AI education platform built by our team that gives deaf children in Pakistan a way to learn — and be understood. Real-time sign recognition through the camera, a 3D signing avatar, and a bilingual English/Urdu curriculum designed for kids aged 4–10.

SignLearn homepage — Every Child Deserves a Voice
"Every child deserves a voice" — the bilingual landing experience.

About SignLearn

Pakistan has one of the world's largest deaf communities, yet Pakistan Sign Language (PSL) resources for children are scarce. SignLearn was built to close that gap: a playful, bilingual learning platform where every child "deserves a voice."

The platform combines a custom-trained AI recognition model with an animated signing avatar named Hashi. Children practice signing letters in front of the camera and get instant feedback; typed or spoken phrases are played back as animated sign language — both directions of the conversation, covered.

Privacy is designed in, not bolted on: camera frames are processed in real time and never stored, and the whole experience is built to child-safety standards with parent and instructor oversight baked into the product.

What we delivered

01

Real-time alphabet recognition

A custom pose-keypoint Transformer model (Keras) recognizes Urdu and English alphabet signs from the webcam in real time, with a language model reranking predictions — 'spell words with your hands.'

02

Speech & text to sign avatar

Spoken or typed phrases are animated as sign language by Hashi, the 3D signing avatar — powered by SiGML sign-notation sequencing running entirely in the browser.

03

Bilingual curriculum & gamification

A structured lesson library over real PSL video clips — greetings, family, numbers, emotions, alphabets — with XP, streaks, levels, and quizzes in tandem English/Urdu throughout.

04

Instructor & parent portals

Instructors build lessons, review community-contributed signs, and track class performance; parents follow each child's progress, practice minutes, and weekly summaries.

05

Accessibility engineering

Bilingual screen-reader support with RTL Urdu, ARIA live announcements for webcam events, reduced-motion support, and kid-friendly feedback in both languages.

Inside the system

SignLearn student dashboard with XP, streaks and sign of the day
The learner's home: XP, streaks, sign of the day, and quiz shortcuts.
SignLearn lessons library with bilingual PSL lessons
The lesson library — 91 signs across bilingual, Urdu, and English tracks.
SignLearn instructor dashboard with lessons, learners and review queue
Instructor space: published lessons, class scores, and the contribution review queue.
SignLearn parent portal with children's progress and activity feed
Parent portal: each child's progress, practice time, and weekly summaries.

System highlights

  • Custom Transformer model for sign recognition (Keras 3)
  • 3D signing avatar with SiGML animation, fully in-browser
  • Bilingual English/Urdu UI with RTL support
  • PSL video dictionary with community contributions
  • Student, instructor & parent roles with progress tracking
  • Privacy by design — camera frames never stored

Outcomes

  • Working AI recognition of Urdu & English alphabet signs
  • Full learning loop: lessons, practice, quizzes, progress
  • Three-sided platform: students, instructors, parents

Built with

ReactFlaskKeras 3MediaPipeJASigning/SiGMLJWT auth
Build something like this

Have a workflow, product idea, or system you want to build?

Let’s turn it into a reliable, scalable software system.

Book a Strategy CallView Our Work